Differentially private stochastic gradient descent algorithm for multiparty classification

We consider the problem of developing privacy-preserving machine learning algorithms in a dis-tributed multiparty setting. Here different parties own different parts of a data set, and the goal is to learn a classifier from the entire data set with-out any party revealing any information about the individual data points it owns. Pathak et al [7]recently proposed a solution to this problem in which each party learns a local classifier from its own data, and a third party then aggregates these classifiers in a privacy-preserving manner using a cryptographic scheme. The generaliza-tion performance of their algorithm is sensitive to the number of parties and the relative frac-tions of data owned by the different parties. In this paper, we describe a new differentially pri-vate algorithm for the multiparty setting that uses a stochastic gradient descent based procedure to directly optimize the overall multiparty ob-jective rather than combining classifiers learned from optimizing local objectives. The algorithm achieves a slightly weaker form of differential privacy than that of [7], but provides improved generalization guarantees that do not depend on the number of parties or the relative sizes of the individual data sets. Experimental results corrob-orate our theoretical findings.

Mechanism design for cost optimal PAC learning in the presence of strategic noisy annotators

We consider the problem of Probably Ap-proximate Correct (PAC) learning of a bi-nary classifier from noisy labeled exam-ples acquired from multiple annotators(each characterized by a respective clas-sification noise rate). First, we consider the complete information scenario, where the learner knows the noise rates of all the annotators. For this scenario, we derive sample complexity bound for the Mini-mum Disagreement Algorithm (MDA) on the number of labeled examples to be ob-tained from each annotator. Next, we consider the incomplete information sce-nario, where each annotator is strategic and holds the respective noise rate as a private information. For this scenario, we design a cost optimal procurement auc-tion mechanism along the lines of Myer-son’s optimal auction design framework in a non-trivial manner. This mechanism satisfies incentive compatibility property,thereby facilitating the learner to elicit true noise rates of all the annotators.

An empirical understanding of use of internal analogies in conceptual design

Internal analogies are created if the knowledge of source domain is obtained only from the cognition of designers. In this paper, an understanding of the use of internal analogies in conceptual design is developed by studying: the types of internal analogies; the roles of internal analogies; the influence of design problems on the creation of internal analogies; the role of experience of designers on the use of internal analogies; the levels of abstraction at which internal analogies are searched in target domain, identified in source domain, and realized in the target domain; and the effect of internal analogies from the natural and artificial domains on the solution space created using these analogies. To facilitate this understanding, empirical studies of design sessions from earlier research, each involving a designer solving a design problem by identifying requirements and developing conceptual solutions, without using any support, are used. The following are the important findings: designers use analogies from the natural and artificial domains; analogies are used for generating requirements and solutions; the nature of the design problem influences the use of analogies; the role of experience of designers on the use of analogies is not clearly ascertained; analogical transfer is observed only at few levels of abstraction while many levels remain unexplored; and analogies from the natural domain seem to have more positive influence than the artificial domain on the number of ideas and variety of idea space.

DTM AT 25: ESSAYS ON THEMES AND FUTURE DIRECTIONS

This paper describes the development and evolution of research themes in the Design Theory and Methodology (DTM) conference. Essays containing reflections on the history of DTM, supported by an analysis of session titles and papers winning the ``best paper award'', describe the development of the research themes. A second set of essays describes the evolution of several key research themes. Two broad trends in research themes are evident, with a third one emerging. The topics of the papers in the first decade or so reflect an underlying aim to apply artificial intelligence toward developing systems that could `design'. To do so required understanding how human designers behave, formalizing design processes so that they could be computed, and formalizing representations of design knowledge. The themes in the first DTM conference and the recollections of the DTM founders reflect this underlying aim. The second decade of DTM saw the emergence of product development as an underlying concern and included a growth in a systems view of design. More recently, there appears to be a trend toward design-led innovation, which entails both executing the design process more efficiently and understanding the characteristics of market-leading designs so as to produce engineered products and systems of exceptional levels of quality and customer satisfaction.

Probabilistic inference for fast learning in control

A novel framework is provided for very fast model-based reinforcement learning in continuous state and action spaces. It requires probabilistic models that explicitly characterize their levels of condence. Within the framework, exible, non-parametric models are used to describe the world based on previously collected experience. It demonstrates learning on the cart-pole problem in a setting where very limited prior knowledge about the task has been provided. Learning progressed rapidly, and a good policy found after only a small number of iterations.